Assessment of four inflow conditions on large-eddy simulation of a gently curved backward-facing step

In this study, we investigated the impact of four different inflow generation methods used in large-eddy simulation on the spatially developing boundary-layer before a recirculation bubble is formed over a gently curved backward-facing step at a Reynolds number 13,700 based on the step height. The configuration under study is very challenging because the separation is caused by a weak adverse pressure gradient, thus making a very sensitive and reliable assessment for evaluating the different inflow conditions used. The first method is a precursor-based simulation in which the velocity data on a certain plane is recorded in a library and then used as the inlet condition for the primary simulation. The second method used is the random noise generation method. The third method is based on generation of turbulent spots which incorporates the distribution of Reynolds stress tensor, and the last one is the so-called rescaling/recycling method proposed by Lund and colleagues. All these methods are compared together in terms of separation and reattachment locations of the recirculation bubble. The flow structures are represented by qualitative criteria, and also streamwise Reynolds stresses and production of turbulent kinetic energy of the flow are assessed and compared together at different stations before and after separation to illuminate how the developing structures within the boundary-layer can affect the locations of separation and reattachment. Distribution of pressure coefficient for different methods showed that there is a relation between production of turbulent kinetic energy and favourable pressure gradient of each method before the separation occurs. Finally, spectra of pressure fluctuation revealed how each inflow condition influences the shedding-instability frequencies.